Over the years, a number of oxymora have established their place among the most popular in the American lexicon-among them "airline food," "jumbo shrimp" and "working vacation."These terms are all regarded seemingly as incongruous or contradictory. So is "business intelligence"--at least in the manner that it relates to the insurance industry.
Simply put, lack of quality data inhibits insurers from using BI tools properly for assessing core areas of their businesses, including risk assessment, underwriting, new product development, claims, and customer service.
"The root problems behind a lack of solid intelligence at many insurance companies can largely be blamed on data quality, incompatible data formats across disparate systems and siloed organizations, or barriers across the organization," explains Matthew Josefowicz, manager of the insurance practice at Celent Communications Inc., a Boston-based research and advisory firm.
"Insurers need to get better at converting the masses of data they have into accessible information so it can serve as operational BI for product line workers or analytical BI for business analysts and product designers. This is where they can examine masses of historical data and determine various trends," he says.
This all sounds well and good, but the problem for insurers hinges on a troubling paradigm: Most are not leaders in championing new technology initiatives. Similar to other IT projects, data mastery is still a work in progress.
"I think you'll find that 10% of insurance companies are leaders (when it comes to data mastery) and another 30% are followers. The rest are laggards who say, 'This is how we've always done things,'" says Lee Fogle, vice president, Insurance Services Office Inc. (ISO), Jersey City, N.J.
According to Framingham, Mass.-based IDC, the market across industries for business intelligence software is worth more than $7 billion worldwide, and that's expected to double by 2006.
And data mastery, which includes business intelligence, ranked among the top three projects that delivered the highest return on investment in a 2003 Celent survey of CIOs and CTOs.
"Two years ago, when we brought up the subject of predictive modeling to an insurance CEO, his eyes would glaze over," says Dax Craig, president, CEO and founder of Valen Technologies Inc., a Denver-based provider of risk-based predictive modeling and analysis tools to the property/casualty market.
"But in a short amount of time, there's been a sea change with insurers about understanding the value of predictive modeling," he adds. "Change management has helped convert some skeptics, but also a dynamic predictive modeling solution offers a high level of transparency, and gives insurers a reason for not only how-but why-a risk is high or low. When insurers see this, adoption goes way up."
Insurers that have perfected data mastery are uncovering compelling trends and behavioral patterns that are contributing to better bottom lines.
For example, Hartford Life Insurance Co. generated record annuity sales in 2003 ($16.4 billion, an increase of 42% from 2002) with the support of technology that was able to detect behavior-from its 1.5 million contract holders and from the stock market.
"We attribute a good portion of the increase in variable annuity sales to the popularity of our income protection guarantee," says Victorio Severino, CIO at Hartford Life, an operating unit of Hartford, Conn.-based Hartford Financial Services Inc.
"Our Principal First product guarantees that the owner of the variable annuity will be able to generate an income from his or her principal investment of up to 7% a year, even if the principal is lost to fluctuations in the equity markets."
The income protection guarantees, however, would be impossible to deliver if the provider lacked a robust, flexible and results-driven system to calculate and assess risk. "An enterprise needs to communicate to business people what drives business intelligence," says Severino. "We're just starting to get an appreciation for the ways we can use data to make it a competitive advantage."
Hartford Life once walked a tightrope in generating sales volume for variable annuity business while insulating itself from risk. A couple years ago, many investors redeemed variable annuities due to stock-market volatility. "The market tanked. We had to give them guarantees to get their investment back, so we created a hedging system-so if the market moved in the opposite direction, we could fulfill these guarantees," he explains.
To support annuity hedging, Hartford Life went live in September 2004 with a grid computing solution to perform complex risk calculations. "Prior to the grid initiative, we had hit a ceiling with the level of horsepower we could deploy in running calculations through servers and desktops," says Severino.
"Now, we use it to support various functions: We use grid computing every quarter to examine the fees we should charge policyholders for services, such as mortality and fund management fees. It's a process that used to take 17 hours to run in an actuarial department, whereas it now it takes 45 minutes."
Developed by researchers at the University of Wisconsin-Madison, the grid computing solution, known as Condor (inspired by the scavenger bird), combines the power of many central processing units (CPUs)--in Hartford Life's case as many as 200 CPUs. The level of risk computation that Hartford must perform requires significant computational horsepower.
"The trouble with the commercial software we used was scalability: 25 CPUs was the threshold," says Severino. "If we needed more CPUs, our IT people would have had to clone software and run it three times to get one result. We could see that we were going to require 200 to 300 CPUs at one point to perform the level of risk hedging needed to grow our business.
"We realized pretty quickly that the grid would bring the stability we needed to deploy significant horsepower," Severino explains.
"When you have CPUs working in unison, if one goes down, you don't have to worry about the whole system crashing, which would negatively impact risk calculation.
"When it came to deciding when we needed to invest in a higher level of analysis, necessity proved to be the mother of invention," he says.
Insurers such as Hartford Life have concluded that leveraging business intelligence tools to track trends, identify various behavior traits and pattern analysis is carried out in vain if information is not delivered in a timely fashion.
For other insurers, timeliness is critical across a number of operational activities, from new business to claims.
"A database might support the tracking of injury claims, treatments and diagnoses-whatever's trending upward," says ISO's Fogle.
For example, insurers might need to reassess whether they're writing an inordinate number of auto policies covering sport utility vehicles, which have a history of rollovers.
But in the process of identifying and reacting, insurers can also pick up other trends such as fraudulent claims. "An insurer might identify a pattern where an attorney works with a physician, and every time they work together the claim is 40% higher than it should be," Fogle notes.
But if insurers uncover these patterns after a claim payment is made-and oftentimes they do-it's extremely difficult to recoup those payments after the fact, such as in an instance of fraud. So timeliness is everything.
The gold standard
"The gold standard is no longer identifying fraud after the fact; we are moving into the realm of software that will do the discovery, and identify the fraud up front," says Rick Pro, vice president of health care informatics, for Pittsburgh-based Highmark Inc., a provider of group and individual health insurance.
"The thing about claims fraud is, once the money is gone, it's hard to get it back," Pro continues. "The return of payment that goes out the door is pennies on the dollar."
In 1998, Highmark formed what it refers to as an "informatics" group to oversee a business intelligence effort across the entire organization. From 1998 to 2001, the informatics group upgraded its hardware, moving analytics to a platform developed by Teradata, Dayton, Ohio. This enabled the group to perform more complex modeling.
The 100-person informatics group is structured independently from Highmark's IT department, but consists of technical programmers and analysts who provide IT support for its business groups.
The informatics group supports such functions as actuarial, finance, medical management, and ad hoc support to a senior management team, says Kristin O'Donnell, the group's manager of product and accounts.
"The industry is very dynamic right now and the focus is on premium dollars and the creation of value," she explains. "Inefficiencies have crept into the process over the years. This has necessitated more dynamic and proactive approaches to medical management."
Using data extraction, data analysis and data manipulation capabilities, informatics provides a high level of analysis that identifies risk factors for various businesses.
"We use risk analysis tools so employers can develop an understanding of what is happening in their own employee pool," O'Donnell explains. "We use predictive modeling to identify high-risk members and to identify if certain (high-risk) members are enrolled in a condition management program."
"Data is the linchpin of everything we do," Pro says. "In the past, we had been limited in interpreting, mining and cleansing our data. We've come a long way since 1998, and in 2005, we plan to move to a new-generation data warehouse that will take us from several weeks lag time for claims adjudication to accessing data within a week of receiving a claim."
Not the panacea
Indeed, insurance companies have a litany of issues that need to be resolved before they can participate in best-practices business intelligence.
One issue has to do with expectation levels. "A lot of people naively think that BI is a panacea for whatever ails their company, and it certainly can be a cure," says David West, who leads insurance industry strategies, for Cary, N.C.-based SAS, a provider of risk modeling and predictive analysis tools.
"But business intelligence is ineffective if insurers have not addressed data quality and data management," he says.
Insurers are also finding they have to trust the results. Just having an underwriter use a scoring solution to arrive at a number to determine whether the company should write the business or not doesn't cut it anymore, say observers.
"Insurers want to understand the rationale behind the scores," says Valen's Craig. "The score might compute to 75, but most line underwriters want to see some level of transparency within the system to indicate why it's 75," he says.
In the final analysis, ISO's Fogle believes the key ingredient to mastering BI is data mastery.
"Insurance companies may have to consider changing their processes because you can't develop a dynamic data warehouse unless you have all the data you need. But all the data you need might still reside in paper claims files," he says.
Hartford Life's Condor System Mines Databases For Gold
Similar to many companies, Hartford Life Insurance Co., an operating unit of Hartford Financial Services Inc., Hartford, Conn., wants to know what its customers are going to do-even before they do it.
And the provider is largely succeeding at that, while also building a formidable variable annuity business that generated record sales in 2003 ($16.4 billion, up 42% from 2002). Practically, the company has turned the volatile variable annuity business into a science by using predictive analysis technology.
Using a grid computing solution known as Condor, which was developed at the University of Wisconsin-Madison, the company has been better able to provide income protection guarantees to customers by using Condor to calculate risk.
Hartford Life's Principal First product guarantees owners of annuities will generate income from their principal investment of up to 7% a year, even if the principal is affected by fluctuations in the equity markets.
But it's not enough to make a guarantee: Insurers such as Hartford Life also have to detect trends to remain profitable. All told, Hartford Life has 1.5 million contract holders.
"We model against these trends and come up with an unexpected risk scenario of the guarantees we offer," says Victorio Severino, CIO at Hartford Life. "We also model risk based on a policyholder's behavior."
With annuity products, investors are notorious for fund activity. Hartford Life believes it's critical to anticipate when this activity will occur. "If I know that Jane Doe always calls Hartford at the end of the quarter to find out about the value of her annuity, then we can anticipate this ahead of time and react," Severino says.
One customer tendency that Hartford Life detects is the tendency to liquidate contracts in favor of other investment options-perhaps outside Hartford Life. The company uses the analysis to retain customers with what could be best described as a pre-emptive strike.
"We can tell when a person might be inclined to shift money from our annuity to another program, based on how their assets have performed," explains Severino.
"They might have held an annuity with us for seven years. We've seen behavior where people will shift money at that point. As a result, we can get on the phone with their broker and make efforts to retain the business."
Hartford Life admittedly has collected a lot of data over the years. "But frankly at one point, we did not know how we were going to use it," says Severino.
"The way we collected data was not unlike a pack rat. But at some point you realize that there's gold in that collection-and you have to mine that gold."
Register or login for access to this item and much more
All Digital Insurance content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access